Using PyMC3

bayesian-inferencemcmcprobabilistic-programmingpython

Abstraction: Hands-on PyMC3 tutorial covering MCMC samplers and Bayesian models

Key points:

PyMC3 supports Metropolis, Slice, and Hamiltonian Monte Carlo (NUTS) samplers for Bayesian inference in Python
Demonstrates convergence diagnostics including Gelman-Rubin statistic (values >1 indicate non-convergence) and Geweke test
Examples progress from coin-bias estimation with Beta-Binomial model through linear regression, robust regression (Student-T likelihood), logistic regression, hierarchical models (Gelman radon dataset), multivariate normal (LKJ prior), and Gaussian mixture models
Traces can be persisted via Text or SQLite3 backends for large models that may not fit in memory
Multi-arm bandit problem solved with Bayesian Beta-distribution updating as a reinforcement learning analog

Connections: Pymc3 · Bayesian Inference · Mcmc · Probabilistic Programming · Hierarchical Models